Git sensei

Your AI guardian for Git. Translate plain English into safe, context-aware commands and code with total confidence.

Comment

1. Inspiration

The inspiration for Git Sensei comes from a universal feeling shared by developers everywhere: a moment of pure anxiety before running a complex Git command. We've all been there—staring at a git reset --hard or git push --force, taking a deep breath, and hoping we don't accidentally wipe out hours of work. Git is incredibly powerful, but its steep learning curve and unforgiving nature can be a source of constant frustration. We wanted to build a tool that could act as a safety net and an expert guide—a "sensei"—that would empower developers to use Git fearlessly and efficiently.

2. What it does

Git Sensei is an AI-powered command-line assistant that makes your Git workflow safer, simpler, and smarter. It acts as an intelligent layer on top of your existing Git installation, transforming how you interact with version control.

  • 🛡️ SAFE: It automatically intercepts potentially dangerous commands before they execute. It explains the risks in plain English and requires your explicit confirmation, acting as a crucial safety net that prevents costly mistakes and data loss.

  • 🗣️ SIMPLE: You no longer need to memorize obscure syntax. You can speak to Git Sensei in plain English. Phrases like "show me the last commit" or "create a new branch called 'feature-x'" are instantly translated into the correct commands.

  • 🧠 SMART: This is the magic. Git Sensei analyzes the context of your repository—like untracked files, your current branch, and recent commits—to provide truly intelligent suggestions. It doesn't just translate your words; it understands your intent, making it a true AI assistant.

3. How we built it

Git Sensei was built from the ground up using Kiro, the AI IDE, with a powerful hybrid development strategy that gave us the best of both worlds:

  1. Spec-to-Code for the Foundation: For the critical core of the application—the safety features, command execution, and overall architecture—we couldn't improvise. We used Spec-to-Code, providing Kiro with our detailed requirements, design documents, and a task-by-task implementation plan. Kiro used this complete blueprint to generate a solid, reliable, and testable codebase. This spec-driven approach was essential for building a tool we could trust.

  2. Vibe for the AI Features: For the creative and exploratory parts, like the natural language translation and context analysis, we switched to Vibe. We had structured conversations with Kiro, using sophisticated prompt engineering to teach our assistant how to be context-aware. Kiro's ability to seamlessly integrate new AI and context modules into the existing architecture, all while handling dependencies and boilerplate, saved us countless hours of work and allowed us to focus on innovation.

4. Challenges we ran into

  • Ensuring AI Reliability: An AI can sometimes be unpredictable. Our biggest challenge was: how do you trust an AI with something as critical as your source code? We solved this architecturally. The Spec-driven safety module was built first and acts as a guardian that checks every command, including those generated by the AI. This means our robust, rule-based system protects the user even from a faulty AI suggestion.

  • Effective Prompt Engineering: Teaching the AI to be "context-aware" was more than just asking a question. It required careful iteration on the system prompt, figuring out the best way to present the repository context so the AI could make logical deductions. This was a challenge we overcame through the rapid, conversational nature of the Vibe development method.

  • The PyPI Name Clash: Our original name, "Git-Gud," was already taken on PyPI. This forced us to pause and rethink our branding. We turned this challenge into an opportunity, choosing "Git Sensei," a name that we feel better captures the project's mission of expert guidance.

5. Accomplishments that we're proud of

  • The Hybrid Development Model: We're incredibly proud of the sophisticated workflow we developed with Kiro, strategically switching between Spec-to-Code for reliability and Vibe for creativity. It proves that human planning and AI-powered generation can work in perfect harmony.

  • The "Wow" Moment: The context-aware AI is our biggest accomplishment. Seeing the application correctly deduce that it needs to run git add when asked to "save my work" because it saw an untracked file was a true "magic" moment. It's the feature that elevates this from a simple wrapper to an intelligent assistant.

  • A Complete, Polished Tool: We didn't just build a prototype. We built a fully functional, well-tested, and documented command-line tool with robust error handling and a professional feel, all within the timeframe of the hackathon.

6. What we learned

  • Plan First, then Vibe: The biggest lesson was the power of our hybrid approach. Building a solid foundation with a clear spec is crucial. Once that's in place, the creative freedom and speed of Vibe can be used to build innovative features on top of a base you can trust.

  • Prompt Engineering is a Superpower: We learned that the quality of AI-generated code and responses is directly proportional to the quality of the prompt. Providing clear context, examples, and constraints to Kiro was the key to unlocking its full potential.

  • AI as a Partner, Not Just a Generator: By the end, Kiro felt less like a tool and more like a development partner. It handled the boilerplate, integrated the modules, and allowed us to focus almost entirely on the core logic and user experience.

7. What's next for Git Sensei

We believe Git Sensei has huge potential to grow. Here are our next steps:

  • Deeper Context-Awareness: We plan to teach the AI to understand more complex scenarios, like branch relationships, merge conflicts, and the difference between local and remote repositories to provide proactive advice (e.g., "You should pull before pushing").

  • Interactive Command Mode: Instead of just executing a command, we want to add an interactive mode where Git Sensei suggests a command, explains what it does, and allows the user to edit or refine it before execution.

  • IDE Integration: While it's powerful on the command line, Git Sensei's true home is inside the IDE. We plan to build extensions for popular editors like VS Code to bring its safety and intelligence directly into the developer's primary workflow.

  • Team-Based Safety Policies: A future version could allow teams to define and enforce custom safety rules across their organization (e.g., preventing force pushes to the main branch).

Built With

Share this project:

Updates